A Worked Example: Calculating Automation ROI on a Real Process
One process, followed end to end, with every number shown. What it cost to run by hand, what automating it cost, what it saved, and how long the payback took.
Most automation business cases are built on an estimate of time saved, produced by asking someone how long something takes. That estimate is reliably wrong, usually low, and it is why so many cases fail to convince a finance director.
This article follows one process end to end with every number shown, including the ones that made the case worse. The method is set out in how to calculate the ROI of automating a business process; this is the method applied.
The business and the process
A UK equipment hire business, 34 staff, £4.1m turnover. The process is the hire return: an item comes back, gets checked, gets marked available, the customer gets a final invoice, and any damage gets charged.
Chosen because it was the most complained-about process in the business, which is a reasonable place to start and not the same thing as the most expensive one.
Step 1: what it actually costs now
Rather than ask, we timed it. Two people, two weeks, recording actual minutes.
| Step | Minutes per return | Who |
|---|---|---|
| Log the return in the hire system | 3 | Yard |
| Condition check and note | 7 | Yard |
| Re-key details into the damage log spreadsheet | 4 | Yard |
| Chase the yard for details when the note was unclear | 3 | Office |
| Produce the final invoice from two systems | 6 | Office |
| Reconcile the damage log against invoices monthly | 90 per month | Office |
Volume: 240 returns a month.
Per-return time: 23 minutes. Monthly: 92 hours, plus 1.5 hours reconciliation, so 93.5 hours.
Fully loaded cost: yard staff at £19 an hour, office at £24, blended to roughly £21.
Current cost: 93.5 x £21 x 12 = £23,562 a year.
Note that the self-reported estimate before timing was "about fifteen minutes". The stopwatch said 23. That gap is normal and it is the argument for timing rather than asking.
Step 2: what else it costs
Invoice errors. 240 returns a month, roughly 3 per cent with a billing error requiring a credit note or a call. That is 86 a year, at about 25 minutes each to resolve plus an average £18 written off. 86 x (£8.75 + £18) = £2,301 a year.
Late availability. Items sat unmarked for an average of 1.4 days after return. On a fleet where utilisation drives revenue, the business estimated 0.4 per cent of potential hire days lost. On £4.1m of hire revenue that is roughly £16,400, and we discounted it heavily to £6,000 because it rests on an assumption rather than a record. Be conservative with anything you cannot evidence.
Total current cost: £31,863 a year.
Step 3: what automating it costs
What was built. A mobile condition-check form used in the yard, feeding the hire system directly. Damage notes and photographs attached to the hire record rather than a separate spreadsheet. Final invoice generated automatically and queued for a one-click office review. Monthly reconciliation removed entirely because the two records became one.
| Item | Cost |
|---|---|
| Build, 14 days | £11,200 |
| Integration licence, annual | £1,440 |
| Training and parallel running | £1,800 |
| Maintenance, 15 per cent of build, annual | £1,680 |
Year one: £16,120. Subsequent years: £3,120.
Amortising the build over three years: £6,853 a year.
Step 4: what it actually saved
Measured, four weeks after go-live, by timing again.
| Step | Before | After |
|---|---|---|
| Log the return | 3 | 0, done by the form |
| Condition check and note | 7 | 6 |
| Re-key into spreadsheet | 4 | 0 |
| Chasing the yard | 3 | 0.5 |
| Final invoice | 6 | 1.5 |
| Monthly reconciliation | 90/month | 0 |
Per-return: 23 down to 8 minutes. Monthly: 32 hours against 93.5.
Time saved: 61.5 hours a month = £15,498 a year.
Errors fell from 3 per cent to 0.8 per cent, saving £1,725 a year.
Availability delay fell from 1.4 days to under 2 hours. Keeping the conservative figure, £6,000.
Total annual saving: £23,223.
Step 5: the ROI
ROI = ((£23,223 - £6,853) / £6,853) x 100 = 239 per cent
Payback: £16,120 year-one cost against £23,223 of saving, so roughly 8.3 months.
What the case did not include
Being straight about this is what makes the rest credible.
The condition check did not get much faster, 7 minutes to 6, because it is a physical inspection. Automation does not help with physical work, and claiming otherwise would have been the easiest way to inflate the figure.
Four weeks of reduced productivity during changeover, not costed, which would reduce year-one return by perhaps £2,000.
Staff satisfaction improved and we did not put a number on it. It belongs in the case as judgement, not arithmetic.
The availability figure is an assumption. It is the least solid number here, which is why it was discounted by nearly two thirds. If you removed it entirely the ROI would be 151 per cent and the payback 11 months, which is still comfortably worth doing. A case that survives the removal of its weakest number is a case worth acting on.
What to take from this
Time the process, do not ask about it. The gap between estimate and stopwatch was 35 per cent here and that is typical.
Discount anything you cannot evidence, heavily. Then check whether the case still works without it.
Measure again afterwards. Almost nobody does, which is why so few businesses can tell whether their automation spending worked.
Include the steps that did not improve. A case where everything got better is a case nobody believes.
Key Takeaways
- Time the process with a stopwatch for two weeks. Self-reported estimates were 35 per cent low here, which is typical.
- Use fully loaded staff cost, roughly 1.3 to 1.5 times salary for UK employees.
- Discount any benefit you cannot evidence, then check the case still works without it entirely. This one did.
- Measure again four weeks after go-live. It is the only way to know whether the money worked.
- Show the steps that did not improve. Physical work does not get faster, and a case where everything improved is not believed.
Frequently Asked Questions
Is 239 per cent a typical return?
For a well-chosen process, yes, and the range is wide. Most worthwhile automations land between 150 and 500 per cent with a payback of three to twelve months. Below about 100 per cent projected, reconsider the process rather than the supplier.
What if we cannot time the process for two weeks?
Time it for three days and accept a wider margin. Even a short measurement beats an estimate. What does not work is asking people, because everyone underestimates routine work they do on autopilot.
Which process should we pick first?
Not necessarily the most complained-about, which is what this business chose. The best first candidate is where information is typed twice, because it is cheap to fix and easy to measure. See 7 business processes every SME should automate first.
The numbers do not work for our process. What then?
That is a valuable conclusion and it is worth recording. If a process consumes under five hours a month with no meaningful error cost, the setup and maintenance will exceed the benefit. Spend the money somewhere it works.
Want this calculation done properly for one of your processes before you commit to anything? Talk to Halo Technology Lab. Our automation solutions start with the stopwatch, and we will tell you when the numbers say no.
Enjoyed this? Get the next one by email
Practical AI playbooks, build logs and tool teardowns. One email a week, free, unsubscribe in one click.
Related Articles
What Does “Automation” Actually Mean in Business?
Not robots, and usually not AI. Automation is getting a computer to do the joining-up a person currently does by hand, and most of it is unglamorous.
What Does “Triage My Inbox” Actually Mean?
Triage is a hospital word: decide what needs attention now, what can wait, and what needs nothing at all. Applied to email it is a simple system, not a piece of software.
Social Media Automation: What to Automate and What to Keep Human
A practical guide to automating your social media without losing authenticity.